Skip to content
Back to skills

Templates

ASecurity

Turnfile protocol execution guide for <AGENT_NAME>

  • 9 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added May 26, 2026
ai-agents

Security analysis

A100/100

Pro scans all 17 files and shows the line behind each finding

Scanned May 27, 2026

npx -y skills add snapsynapse/turnfile --skill templates --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Templates?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Templates
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/snapsynapse-templates/badge)](https://www.skillsdirectory.com/skills/snapsynapse-templates)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: turnfile-<agent>-collaboration
description: "Turnfile protocol execution guide for <AGENT_NAME>"
---

# Protocol Skill — <AGENT_NAME>

<!--
Copy this template to skills/<agent>/SKILL.md and fill in agent-specific details.
This file is the agent's complete, self-contained protocol execution guide (PRD-012 R1).

Key principles:
- Each agent maintains its own full skill file (no shared core with thin wrappers).
- The repository copy (skills/<agent>/SKILL.md) is canonical.
- Policy tests in skills/policy-tests/ validate cross-agent consistency.
- Skills are invoked only by explicit maintainer instruction (PRD-012 R3).
-->

## Overview

| Field | Value |
|-------|-------|
| Version | 0.1.0 |
| Status | Draft |
| Protocol baseline | PRD-003, 004, 008, 009, 010, 011, 013, 014 |
| Last updated | YYYY-MM-DD |

## How to use

Maintainer invokes modules by name. This skill file is **not** self-triggering.
The agent reads this file at session start for operational guidance.

## Execution contract

1. Read TURNFILE.yaml for current coordination state.
2. Execute the requested module steps.
3. Propose changes — do not write without explicit maintainer confirmation (PRD-012 R5.1).
4. Log all substantive actions in WORKLOG and/or MAILBOX.

## Active-turn boundary discipline (PRD-012 R2.2)

1. **First step:** Read mailbox inbox snapshot; check for unread messages.
2. Execute assigned work.
3. **Last step:** Re-check mailbox; confirm unread = 0 before declaring turn complete.
4. If unread cannot be cleared, escalate to maintainer with blocker context.

## Startup orientation read order (PRD-011 R3)

1. `working-session/boot-<agent>.md` — this agent's boot file
2. `working-session/TURNFILE.yaml` — coordination state
3. `working-session/WORKLOG.md` lines 1–11 (status block)
4. `working-session/MAILBOX.md` inbox snapshot
5. `working-session/OPEN_QUESTIONS.md` Active + Deferred sections
6. `working-session/chat-<agent>.md` session close snapshot (bottom of file)
7. Files relevant to current task

## Modules

<!--
Each module corresponds to a protocol workflow (PRD-012 R2).
Define: trigger phrases, expected inputs, deterministic outputs, stop/escalation conditions.

Example module:

### Module 1 — Session start (PRD-011, PRD-013)

**Trigger phrases:** "start session", "boot up", "initialize"
**Inputs:** boot file, TURNFILE.yaml
**Steps:**
1. Read boot file for orientation.
2. Read TURNFILE.yaml for coordination state.
3. Read WORKLOG status block.
4. Read MAILBOX inbox snapshot.
5. Read OPEN_QUESTIONS Active + Deferred.
6. Read chat-<agent>.md session close snapshot.
7. Report readiness to maintainer.
**Outputs:** Agent status set to `active` in TURNFILE.yaml (pending confirmation).
**Stop condition:** All orientation reads complete and reported.
**Escalation:** If boot file is missing or stale, notify maintainer.
-->

### Module 1 — Session start (PRD-011, PRD-013)

### Module 2 — Mailbox lifecycle (PRD-003, PRD-012)

### Module 3 — Maintainer decision request (PRD-004)

### Module 4 — Cross-agent review (PRD-008, PRD-009)

### Module 5 — Shared-file transaction (PRD-010, PRD-013)

### Module 6 — Session close (PRD-011, PRD-013, PRD-014)

### Module 7 — Turnfile coordination (PRD-013)

### Module 8 — OQ registry management (PRD-009)

## Fallback rules

<!--
Define graceful degradation when tools or artifacts are unavailable.
Example:
1. If TURNFILE.yaml is missing, notify maintainer and operate in propose-only mode.
2. If mailbox validator is unavailable, perform manual inbox count.
-->

## Output format requirements

<!--
Define how the agent reports results to the maintainer.
Example:
1. Summarize actions taken and files modified.
2. List any unresolved items or blockers.
3. Provide revision token for any modified artifacts.
-->

## Environment-specific notes

<!--
Agent-specific tooling, context window management, or behavioral notes.
These may differ between agents without affecting protocol outcomes.
-->

## Versioning

| Field | Value |
|-------|-------|
| Skill version | 0.1.0 |
| Validated against | — |
| Policy test status | not yet run |
| Aligned with PRDs | 003, 004, 008, 009, 010, 011, 013, 014 |

Files in this skill

  • SKILL.md4.1 KB
  • counter-recommendation.md963 B
  • decision.md583 B
  • handoff.md779 B
  • prd.md3 KB
  • proposal.md684 B
  • retrospective.md1018 B
  • session-charter.md1.7 KB
  • working-session/MAILBOX.json462 B
  • working-session/MAILBOX.md1.5 KB
  • working-session/MAILBOX_ARCHIVE.md663 B
  • working-session/OPEN_QUESTIONS.md1.1 KB
  • working-session/TURNFILE.yaml2.2 KB
  • working-session/WORKLOG.md1.4 KB
  • working-session/WORKLOG_ARCHIVE.md589 B
  • working-session/boot-agent.md3.3 KB
  • working-session/chat-agent.md1 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.